Applying Eigen-enabled Cognitive Machine Technology to Stop Heart Attacks -- Testimonials Title
Dr. John Showalter and the University of Mississippi Medical Center are using Jvion's Acute Myocardial Infarction (AMI) predictions to stop heart attacks. Jvion's solution is performing almost two times better than stress tests and 20% better than CT coronary angiograms in predicting AMI events in a low risk population within 12 months of discharge.
"Smack dab" in the middle of Mississippi sits the University of Mississippi Medical Center and their Chief Health Information Officer John Showalter. Like most academic facilities, they treat some of the most complex cases and sickest patients. But in Mississippi things are even more challenging.
Dr. Showalter and his team treat patients who are more likely to smoke, be overweight, and to have had a stroke. They are also least likely to exercise or to have health coverage. Not surprisingly, heart disease is the leading cause of death in Mississippi.
This reality sat heavy on Dr. Showalter's mind. One out of every 10 patients leaving UMMC suffered a heart attack within 12 months. 'This has to stop' thought Dr. Showalter, 'there simply has to be a better way. If we could just predict patients at risk of a heart attack within a year, we could better allocate resources and improve the health of Mississippians.'
So Dr. Showalter looked to Jvion. He knew that the predictive engine within the firm's software was extremely powerful and he knew the Jvion team. But could a custom use case be developed that would target Acute Myocardial Infarctions (AMIs)? And could we stand up the solution quickly?
Dr. Showalter brought together his team of experts to work with Jvion. Their goals were lofty:
- Tune a use case to accurately predict AMIs
- Use the data UMMC had on hand – no matter how dirty or incomplete
- Integrate the solution into UMMC’s Epic instance
They got to work. Jvion used 835 and 837 data along with lab and medication order data to build the machine learning model. UMMC worked to validate the predictions, provide preferences, and deliver data support when needed.
After just a few months tuning the engine and integrating the predictions into Epic, Dr. Showalter and UMMC can take the entire population of patients and zero in on a target group of 1.5%.
Within that group, 75 out of 100 patients are predicted to have a heart attack within 12 months. Since the implementation of UMMC’s solution, more than 10,000 patients have had more than 50,000 daily predictions automatically rendered through the EHR.
'This solution has helped us identify a previously unidentifiable population—patients at risk of an AMI in the next 12 months. This population will require new and unique interventions. And their identification will allow us to explore the most effective interventions to improve outcomes for this unique group.' – Dr. John Showalter
In the coming year, UMMC expects to monitor more than 25,000 people and track the outcomes of interventions on these high-risk patients."